Turn Taking Phrases

Pragmatics 16:2/3.329-360 (2006)

International Pragmatics Association

SPONTANEOUS AND NON-SPONTANEOUS TURN-TAKING

Maite Taboada

Abstract

Turn-taking is usually considered to follow a simple set of rules, enacted through a perhaps more

complicated system of signals. The most significant aspect of the turn-taking process is that, in most cases,

it proceeds in a very smooth fashion. Speakers signal to each other that they wish to either yield or take

the turn through syntactic, pragmatic, and prosodic means. In this paper, I explore how the turn-taking

process develops in two different sets of Spanish conversations. In the first group of conversations,

speakers take turns spontaneously, presumably as they would do in everyday situations. In the second

group, turns were mechanically controlled, and communication was one-way. A comparison of the two

types of conversation provides insights into the signals used in spontaneous turn-taking.

Keywords: Conversation, Turn-taking, Task-oriented conversation, Spanish.

1. Taking turns when talking

Goodwin (1981: 24), reporting on a comparison by Jaffe and Feldstein (1970), proposes

that everyday conversation is similar to short-wave radio as to how the turn-taking is

performed. The speaker provides an end-of-message signal, after which the hearer holds

the channel, bringing about a change in the speaker/hearer roles. In one-way short-wave

radio communication, this end-of-message signal is verbalized in a pre-established

word, in English usually “over”. The difference between the two types of interaction is

that, in a normal conversation, speakers avail themselves of other means or mechanisms

to provide that end-of-message signal. My purpose in this paper is to explore which

exactly are those mechanisms that speakers use in order to signal turn-taking, with a

focus on Spanish.

This study is an analysis of a corpus that contains conversations between dyads of

native speakers of Spanish. The conversations are task-oriented - participants have to

agree on scheduling an appointment - and were recorded using two different procedures.

In a subset of the conversations, the speakers used a computer to indicate the end of a

turn. When they had finished their turn, they press the “Enter” key on a keyboard to

330 Maite Taboada

yield the floor to their interlocutor. This group of conversations is comparable to

conversations via one-way communication channels. In the other subset of the corpus,

turns are spontaneous, without the need for mechanical intervention. In neither case did

speakers establish visual contact, making it impossible to communicate turn-taking

through gaze or other gestures. As a consequence, in the second group of conversations,

the speakers must supply other verbal signals in place of the “over” signal. The signals

may range from pauses (filled or unfilled) or discourse markers to questions addressed

at the interlocutor.

In this paper I describe the analysis carried out in both types of conversation. The

focus of attention is on turn-constructional units (Sacks et al. 1974), and the ending and

beginning portions of a turn (the turn-assignment components). Through an examination

of the resources available to speakers, I establish a preliminary description of how

Spanish speakers signal turn-taking in spontaneous conversation. The corpus used for

this study happens to be a most appropriate tool to test Goodwin’s observation. The

recording situation is exactly the same for both types of conversations, the only

difference being the mediating of the turn-taking through a computer keyboard. Section

3 describes the data collection and other characteristics of the corpus. The next section

reviews previous studies of turn-taking mechanisms, and Section 4 examines the

characteristics of turn-taking and turn-yielding in the corpus. Section 5 ends the paper

with a discussion of the phenomena observed.

2. Studies in turn-taking behaviour

Sacks, Schegloff and Jefferson (1974) observed a number of characteristics in

conversation, among them: Variable turn order and size; variable distribution of turns;

overlapping is common, but brief; and overlapping is promptly repaired (when two

parties find themselves speaking at the same time, one of them will stop). Given these

characteristics, it is obvious, according to Sacks et al., that turn-allocation techniques are

being used. The current speaker may select a different next speaker, or either party may

self-select. This paper is concerned with those turn-allocation techniques that are

“obviously” present.

Sacks et al. (1974) propose a set of rules that apply at each transition-relevance

place, that is, at the point where a next turn can be expected. At each transition-

relevance place, choices are presented to both speaker and hearer(s) as to who is to utter

the next turn-constructional unit. However, Sacks and his colleagues did not detail the

signals employed in communicating such choices. Their rules only specify that, at any

given transition-relevance place, the turn-so-far might be “so constructed” as to involve,

or not to involve, the use of speaker self-selection. In summary, turn-taking consists of a

set of rules, which are enacted through the use of signals (Duncan 1972, 1973).

Spontaneous and non-spontaneous turn-taking 331

Here, I refer to a turn, and to turn-taking, as instances of floor-taking with the

intention of holding the floor for a certain period of time. Thus, a turn is different from

the situation where a speaker produces backchannel signals (Yngve 1970). Backchannel

signals, such as uh-huh, right, yeah, etc., are signals that the channel is still open, and

they indicate at the same time that the listener does not want to take the floor. Duncan

(1972) also establishes a distinction between simultaneous turns and simultaneous

talking. Instances of the first involve true overlapping, whereas instances of

simultaneous talking do not always imply that the current hearer intends to take the turn;

they might just be the result of backchannel signals overlapping with the current

speaker’s turn.

In the rest of this section, I discuss the nature of a turn, and the different possible

signals and rules that have been proposed to account for the observed turn-taking

behaviour in conversation.

2.1. What is a turn?

According to Edelsky (1981), turn definitions can be grouped in two main camps:

Mechanical and interactional (see also Furo 2001). The first group treats turns as units

of talk in interaction, without taking into account social context. In this group are studies

by Jaffe and Feldstein (1970) and Duncan and Fiske (1985), because for them the turn

merely consists of talk with an end boundary. Turns are attributed to a single speaker

and are defined in terms of the behaviour of other parties in the conversation (a turn

ends when somebody else claims the floor). Goffman (1981) says that a turn is the

opportunity to hold the floor, not necessarily what is said while holding it.

On the other hand, interactional definitions are concerned with what happens during

the interaction, and take into consideration the intention of the turn taker. Edelsky

(1981) points out that speakers are more concerned with completing topics than

structural units. Therefore, she defines turn as instances of on-record speaking, with the

intention of conveying a message. She also differentiates turn and floor, since it is often

difficult to determine who has the floor, such as situations where a turn is constructed

collaboratively by more than one speaker. The floor is the activity taking place or the

topic being discussed, often done in collaboration. Hayashi (1991) expands on this

definition of floor, describing it as a means of orientation to the communication at the

higher level of conversation structure.

Selting (1998, 2000) carries out an extensive review of the meaning behind the

concept of turn-constructional unit (TCU) as a unit of talk. TCUs were proposed by

Sacks et al. (1974) as the basic units of conversation. Each TCU ends in a transition

relevance place, that is, the place where the turn may shift to another speaker. Selting

characterizes the notion of TCU as holistic and in need of interpretation. Study of TCUs

332 Maite Taboada

is relevant here, because one TCU may constitute a complete turn. TCUs may be as

short as a word or as long as a sentence. Selting discusses the criteria to divide a turn

into units, and concludes that it requires examination of both syntactic and prosodic

components.

As we will see in the next section, units in conversation are defined by their

boundaries: A unit is talk produced up until an end-point, the point where another

interlocutor can take the floor. Ford and Thompson (1996) defined units as those

characterized by ending in a complex transition relevance place. They added “complex”

to Sacks et al.’s (1974) transition relevance places, because turn units were found to be

identifiable through the complex interaction of syntactic, prosodic and pragmatic

signals. One interesting aspect of their study is that they used backchannels and laughter

by the interlocutor as a signal of a possible completion point of the current speaker’s

turn. Backchannels by the interlocutor are produced at a point where the current speaker

could finish their talk, i.e., a (complex) transition relevance place. That is certainly the

case: Backchannels are produced at a point where an interlocutor could take the turn, but

with the backchannel the interlocutor signals that they do not want to do it, in addition

to signalling their understanding of or agreement with what is being said (Schegloff

1982; Yngve 1970). Ford et al. (1996) also present an extensive discussion on the

definition of a unit in talk, concluding that TCUs are emergent, rather than pre-defined,

and that syntax, prosody and gesture all contribute to defining the basic unit of talk.

In this paper, I define a turn as continuous talk by one speaker, uninterrupted by the

other speaker. There may be talk by the other speaker, but that is often in the form of

backchannel signals, which do not constitute instances of turn change. In Example (1),

speaker FJCD continues her turn after speaker FLNB has uttered a ya (‘I see’). This

backchannel is delivered in overlap with part of speaker’s FJCD turn, which continues

after the backchannel. Therefore, the two parts constitute a single turn1.

(1) Cross-talk

fjcd_07a: déjame ver. yo la semana que entra tengo una reunión de nueve a once el lunes,

/mm/ y después <voy a est> <voy a> voy a estar de viaje. el /uh/ martes miércoles y jueves.

o sea el nueve diez y once. [ entonces ]

flnb_08: [ ya ].

fjcd_07b: /ah/ <tal vez pode> <podríamos> bueno yo tengo <s> libre el viernes, de once a

una.

‘Let me see. Next week I have a meeting from 9 to 11 on Monday, mm and then <I’m

going to b> <I’m going> I’m going to be away. On uh Tuesday, Wednesday and Thursday.

That’s the 9th, 10th and 11th.’

‘I see.’

1 See Section 3 for a description of transcription conventions.


‘Uh <maybe we ca> <we could> well I have <s> free Friday, from 11 to 1.’

2.2. Turn-taking signals

The study of conversational organization has always assumed that there are ways in

which speakers communicate the desire to yield, take, or maintain the floor. Sacks et al.

(1974) assumed that such signals exist, although they did not discuss any particular

signal. Later research has concentrated on a number of different signals: Discourse

markers, pauses and silence, pitch, and intonation. Of importance are also the syntactic

and semantic characteristics of the turn so far, i.e., whether the message can be

constructed as a complete one from a syntactic or semantic point of view.

Duncan (1972) proposed that, in every interaction, there are signals that speakers and

hearers send to each other in order to indicate their state with regard to the turn. Turn-

yielding signals include: Intonation (rising or falling pitch); drawl; body motion

(termination of hand gesture, relaxation of a tensed hand position); “sociocentric

sequences” (fixed expressions such as or something, you know, but uh); paralanguage

(drops in pitch or loudness); syntax (complete grammatical unit)2.

Sacks et al. (1974) consider that syntactic information is important. A complete turn-

constructional unit is one that can be interpreted as a syntactic unit, whether a sentence,

clause, phrase or word. Each one of those units has a component of projectability: The

interlocutor knows that the unit is possibly complete from a syntactic point of view.

Content and genre play a role in projectability of a completed unit. Selting (1998)

points out that the pre-sequences that introduce narratives are licenses to talk for a

relatively long period of time (see also Houtkoop and Mazeland 1985; Sacks 1992).

Similarly, when telling jokes, the speaker may preface the contribution in a way that

makes his or her interlocutor(s) suppress the desire to take the floor (Have you heard the

one about…?).

Intonation is fundamental in the interpretation of talk. Chafe’s (1994) intonation units

are defined as basic segments of talk interrupted by the human need to breathe.

Intonation units are characterized by changes in pitch (fundamental frequency),

duration, intensity, and alternation of talk and silence (pauses). A number of studies

examine the pitch characteristics that signal the end of a turn. Beattie et al. (1982)

analyzed an interview with Margaret Thatcher, and determined that she signalled the

end of a turn when she, however, did not intend to yield the turn. At the interrupted

points, she has a fast pitch fall similar to that in her turn-final utterances. This led to

2 Signals and correlations between signals and turn-taking are further elaborated in Duncan

(1973) and Duncan and Fiske (1977).

334 Maite Taboada

frequent interruptions, because the interviewer interpreted the pitch change as a turn-

yielding signal. Similarly, in a study involving subjects that either read a transcript or

listened to an audio recording of both turn-medial and turn-final utterances, Stephens

and Beattie (1986) showed that subjects could identify the turn-final utterances only if

presented with an audio recording. Cutler and Pearson (1986) found that there are a few

contours that indicate turn-yielding. Whatever the characteristic, it is clear that

phonological information is at play.

Silence and hesitation markers (unfilled and filled pauses) are also signals for

interlocutors (Maclay and Oswood 1959). Beattie (1977) found that people were

interrupted more often during a silence (unfilled pause), and that filled pauses tended to

follow unfilled pauses. In other words, when a speaker does not produce any talk

(unfilled pause), but still wants to hold the floor, a filled pause is produced, to signal the

desire to continue talking. Drawl and perceived duration may also play a role in

distinguishing turn-medial from turn-final utterances (Stephens and Beattie 1986).

Ford and Thompson (1996) found that pauses helped identify completed intonation

units, the minimal noticeable pause being 0.3 seconds long. However, pauses are not

always indicators of an intention to yield the floor. Local and Kelly (1986) proposed that

pauses are of two different types: One that signals the intention to keep the floor (a

‘holding’ silence), and another that indicates that the interlocutor may claim it (a ‘trail-

off silence’). They particularly examined pauses preceded by a filled pause (uh, um, so),

following Jefferson (1983). Local and Kelly distinguished the two types of silences

phonetically: In holding silences there is a glottal closure after the filled pause,

maintained through the silence, and released at the beginning of the following word by

the same speaker. Trail-off silences have an out-breathing at the end of the filled pause,

which also has a more centralized vowel.

Gaze and gesture are also an important signal in the management of turns. It is clear

that gaze can be a turn-allocating mechanism, and that it is an integral part of face-to-

face conversation (Beattie 1979; Lerner 2003). Kendon (1994, 2002) has shown that

gesture fulfills a variety of purposes, among them those typically performed by

discourse markers (Kendon 1995). Speaker and listener movement serve as signals for

turn-taking: Termination of a hand or arm gesture signals the desire to yield the turn,

and continued gesticulation by the speaker acts as a signal to suppress turn-taking by the

hearer, similar to the effect of a filled pause (Beattie 1981; Duncan 1972). However,

since my study consisted of conversations with no face-to-face interaction, I will not

discuss gaze and gesture further.

A combination of features, rather than one individual item, is probably the best

indicator of turn boundaries. Duncan and Fiske (1985) present a number of studies that

examine the interaction of variables such as gaze, gesture, filled pauses, and the

structure of adjacency pairs. Ford and Thompson (1996) studied the interplay of

syntactically complete units, intonation and pragmatic closure, and found that, given a

syntactically complete unit, it is the combination of intonation (marked fall or marked

high rise in pitch at the end of the intonation unit) and pragmatic completion (the unit is

interpretable as a complete conversational action) that most often signals a possible turn


shift, the complex transition relevance place. Wennerstrom and Siegel (2003) also see

turn-taking as a complex process, possible through the interaction of both phonological

and syntactic cues. In their study, they examined the interaction of intonation, pauses,

and complete syntactic units, concluding that it is a complex interaction of the three that

indicates that a speaker’s turn has ended, and the floor is open. In some cases, intonation

overrides syntax, and it can signal turn continuation despite a syntactic boundary. The

intonation pattern with the highest likelihood of indicating a turn shift was the high rise

(H-H% in Pierrehumbert’s (1980) model), although low rise (L-H%) was also found to

indicate turn shift. The most interesting aspect is that not all the high rise utterances

were questions from a syntactic point of view. They found that, when longer pauses

were produced (0.5 seconds), the current speaker resumed talk. This could be because

the opportunity for the other speaker to take the floor, at 0.3 seconds (Ford and

Thompson 1996), had been missed, and the current speaker decided to continue talking.

It is also worth noticing that Wennerstrom and Siegel found speaker differences in

preference for keeping the floor.

Most of the studies cited above focus on English turn-taking. There are a few studies

on prosodic and other signals for turn-taking in languages other than English. For

instance, Auer (1996) examines syntactic and prosodic cues in combination in German

conversations, and there are a few studies on Japanese (e.g., Hayashi 1991; Tanaka

2001). In Spanish, Placencia (1997) points out that in Ecuadorian telephone

conversations, closings are very similar to those described by Schegloff and Sacks in

English (1973), but that cultural factors account for differences in face-saving strategies

(Brown and Levinson 1978). Hidalgo (1998) examines prosody and its relationship to

turn-taking.

In this study, I am also interested in the function of discourse markers as signals in

the turn-taking process. By discourse markers I mean a varied group of conjunctions,

interjections, filled pauses, adverbs and adverbial phrases, such as okay, yeah, right, uh-

huh, and, so, I mean. The trouble with discourse markers is that they do not fit one of

the three basic types of signals: Syntactic, semantic or intonational. They are certainly

not prosodic in nature (although they may have their own prosodic characteristics), but it

is difficult to say whether they contribute syntactic or semantic information to determine

whether the turn is ending, and whether the interlocutor desires to take the turn.

Wennerstrom and Siegel (2003), for instance, classify discourse markers together with

other syntactic devices that may help in turn-taking.

Discourse markers are generally assumed to signal relations among propositions or

among sentences (Fraser 1999; Knott and Sanders 1998; Taboada in press, and many

others); they serve to link global and local discourse structure (Redeker 1990; Schiffrin

1987), or to indicate a return to a previous topic after a digression (Grosz and Sidner

1986); they indicate a dispreferred second part in an adjacency pair (Pomerantz 1984;

Schiffrin, 1987); they can serve as acknowledgment tokens (Jefferson 1984) or

backchannel signals (Yngve 1970); help listeners integrate information in spontaneous

talk (Fox Tree and Schrock 1999); or monitor the interlocutor’s comprehension of the

speaker’s meaning (Fox Tree and Schrock 2002). Bangerter et al. (2004) examine what

336 Maite Taboada

they call project markers (uh-huh, yeah, right, okay) as signals of transitions between

different parts of a telephone conversation. In their analysis, those words have a function

at the global level of organization of the conversation (and the task), rather than at the

local level of turn-taking.

Sacks et al. (1974: 719) discuss turns that begin with an appositional beginning, such

as well, but, and, so. The appositional beginnings give the speaker some time to think

about uttering a complete sentence, and, if there is overlap, make it possible for the

interlocutor to capture the full utterance from its “real” beginning (i.e., uttering one of

these means: “I’m about to start talking, listen up”). If the previous speaker is not yet

listening and missed one of these appositional beginnings, they did not miss any

important content. They call them turn-entry devices, or “pre-starts”. They propose that

these devices be understood as “devices with important turn-organizational uses” (Sacks

et al. 1974: 720). These turn-initial (or TCU-initial) signals are also called filled pauses

(Beattie 1977; Maclay and Oswood 1959).

A few individual discourse markers have been studied as to their role in turn-taking:

Schiffrin (1987) reviews the role of turn-initial well, and, so, and but, showing

differences in the content of the turn they start (cooperative, continuative, contrastive).

Condon (1986, 2001) has extensively studied the discourse functions of ok, among

which is the marking of boundaries in decision-making processes (i.e., if not directly a

turn-taking device, ok helps mark that one portion of the discourse is over, and thus the

floor may be open). She also discusses other studies of ok that have pointed out its role

as closure in phone conversations (Schegloff and Sacks 1973); as a marker of a topic

transition (Schegloff 1968); or as a marker that the conversation is proceeding as

expected, as opposed to well, which may signal a dispreferred second part in an

adjacency pair (Schiffrin 1987). A similar function is achieved in Spanish with pues

(Serrano 1995) or bueno (Briz 1993).

Redeker (1991) reviews some of Schiffrin’s (1987) functions, and adds a few more,

among them the function of certain discourse markers in turn-taking: Oh can elicit a

clarification question (thus allocating the turn back to the previous speaker)3; now and

then can be used to keep the floor; I mean is sometimes a device for starting a turn.

The marker so is also thought to be involved in turn-taking, since it serves as a

marker of a summary (upshot) of what has previously happened, and can therefore

signal the last unit in a turn (Raymond 2004). Raymond points out that so can initiate a

unit that is meant to be the upshot, or it can be the upshot itself (without a full unit

following so).

Ferrara (1997) examines various uses of anyway. Among them is the expression by

the speaker that he or she wants to regain the floor, after an interruption (although

mainly it serves to manage digressions by the speaker).

3 On other uses of oh, see Local (1996), Heritage (1984), and Schiffrin (1987).


In summary, the most extensively studied signals for turn taking are intonation,

silence, filled pauses and discourse markers. In Section 4 I explore how the last three

contribute to maintaining and claiming the floor in a particular corpus of Spanish

conversations. Intonation, although an important factor, is not included in the present

study. The next section provides further information on the corpus used.

3. The corpus

The corpus consists of 60 conversations between dyads of two speakers each, 30 of

which are push-to-talk and 30 cross-talk. These terms refer to the recording style. In

push-to-talk, speakers have to push the “Enter” key on a computer keyboard in order to

gain the floor. In cross-talk, the turn-taking is not impeded by any mechanical means,

i.e., it is as natural as in a telephone conversation (Hopper 1992). Within each group, the

conversations are broken down in ten female-female, ten male-male, and ten female-

male dialogues. They were chosen to obey, besides the gender balance, two other

constraints: That they be approximately the same in length, and that as many different

speakers as possible be represented (in the recording of the large corpus, speakers often

recorded more than one conversation).

The dialogues were collected by the Interactive Systems Laboratory of Carnegie

Mellon University as part of JANUS, a speech-to-speech machine translation project4.

The two recording styles were tested as part of experiments to determine which type

would be easier to process and translate. The speakers were recruited and brought to a

lab for the recording. Some of them knew each other beforehand; some of them were

introduced at the beginning of the recording session. The instructions explained that the

participants have two conflicting agendas (provided to them by the researchers) covering

a period of two to four weeks. The participants need to agree on a two hour appointment

within that time frame. Further details on the corpus are provided in Taboada (2004).

Speakers were mostly undergraduate and graduate students at Carnegie Mellon

University or the University of Pittsburgh. They came from all corners of the Spanish-

speaking world: Argentina, Chile, Colombia, Costa Rica, Dominican Republic, Mexico,

Peru, Spain, Venezuela, including some raised in the United States (Florida, New York).

It is quite possible that there are dialectal variations in the phenomena studied here. My

claims are to be interpreted as general with respect to Spanish; further research could

detail what aspects are specific to each dialect.

4 Thanks to the Interactive Systems Laboratory and to Alex Waibel, its director, for permission to

use the corpus. The selection used here is part of a larger corpus of approximately 500 Spanish

conversations, and a number of conversations in other languages.

338 Maite Taboada

The dialogues were transcribed by members of the JANUS project. The transcripts

include a number of conventions introduced by the transcriber, to reflect every sound

produced during the conversation. For ease of reading, I have deleted most of them, only

leaving some that seemed important: /hm/ and /um/ indicate a hesitation on the part of

the speaker. Stretches of talk accompanied by laughter are surrounded by /begin_laugh/

and /end_laugh/. False starts or repetitions are indicated with < > angled brackets

surrounding the material that was repeated or repaired. Overlaps are indicated with

square brackets [ ] around the words that are produced at the same time. Backchannels

and filled pauses are marked with forward slashes: /uh/. Pause length is indicated in

parentheses. Turns are indicated through the initials of the speaker, which always start

with either an “f” (for females) or an “m” (males), plus the person’s initials, followed by

the turn number. Each example indicates whether the source is part of the cross-talk or

the push-to-talk corpus.

In addition, transcriber comments include intonation, marked with a comma (,), a

period (.) or a question mark (?) at the end of the corresponding section of speech. These

markings do not reflect, nor are influenced by, sentence structure. The speaker may have

the intonation of a statement when he or she is, in fact, asking a question. He or she may

have the falling intonation typical of the end of a sentence (reflected in a period) after a

collection of words that do not, in any way, resemble a grammatically correct or

complete sentence. For this reason, the word after one of these intonation markers is not

usually capitalized, to avoid confusing them with regular periods or question marks.

The translation of Spanish examples is a free translation, rendered one clause at a

time. I have tried to capture the overall meaning, rather than translating word-by-word,

but staying close to the original, which may make the English translations sound

awkward in some cases. When false starts are produced, they sometimes indicate the

beginning of a word or phrase that was never completed. It is possible sometimes to

guess what the word would have been. In those cases, the English translation reflects a

close translation of the interrupted word. In cases where it was not possible to determine

the word intended, the false start sound is the same in the translation as in the original.

The conversations are considered instances of a genre, as a purposeful, staged, goal-

oriented activity (Martin 1984), which I have called scheduling genre (see also Taboada

2004). They are instances of talk produced for a very specific purpose, that of setting up

an appointment. As a result of their practical purpose, the conversations are staged in

particular ways: One speaker proposes a meeting, perhaps also a time; the other speaker

replies either with a different time or with their availability for the time proposed. The

conversation continues until a day and a time have been set. As a result of their social

function, the conversations usually have Opening and Closing stages, and polite devices

that will avoid face-threatening acts to the other speaker. The task has a clear structure,

consisting of steps that have to be taken in a certain sequence in order to successfully

complete the task.

Tables 1 through 3 provide further information on the corpus. Table 1 lists the

number of turns, units and words in both subsets of the corpus. It shows that the cross-

talk conversations contain roughly twice as many units and words as the push-to-talk


conversations. With regard to turns, the cross-talk conversations have more than three

times the number of turns, as compared to the push-to-talk conversations. In other

words, speakers in cross-talk conversations say about twice as much as the other

speakers, but use almost four times the number of turns.

Push-to-Talk Cross-Talk

Turns 248 931

Units 1294 2319

Words 9112 18470

Table 1. Total number of turns, units and words

These tendencies are more clearly presented in Table 2, which provides the average

length of turns, units and words, and also the ratio of words to turns (on average).


Average length - turns 8.27 31.03

Average length - units 43.10 77.30

Average length - words 225.93 315.53

Ratio words : turns 27.32 10.17

Table 2. Average lengths per dialogue

Table 3 breaks down the average length according to the speakers’ gender. In general,

the male-male conversations seem to be longer within the push-to-talk environment. In

cross-talk, it is male-female conversations that are longer under all accounts (turns, units

and words). It is difficult to establish whether this is a general tendency, or simply an

accident for the ten conversations included in each group.


F-F M-M M-F F-F M-M M-F

Average length - turns 8.2 8.4 8.2 31.4 26.9 34.8

Average length - units 39.1 49.5 40.7 75.4 69.7 86.8

Average length - words 291.7 342.5 273.1 305.4 302.3 338.9

Table 3. Average lengths per dialogue, by gender. F=female, M=male

340 Maite Taboada

4. Turn taking in task-oriented dialogue

The management of floor in specific activities may be slightly different than in

spontaneous conversation, the type of interaction that has been described in most studies

mentioned so far. For instance, Jones and Thornborrow (2004) describe floor in

classroom activities as a result of the activity at hand. They characterize floor “as

something people participate in […], rather than “hold”.” (Jones and Thornborrow

2004: 420). In this sense, the concept of floor is similar to Edelsky’s (1981)

collaboratively developed floor. It is important to note that Edelsky also observed a

specific activity, administrative meetings at a university.

In general, the conversations make turn-taking necessary at certain times, because

they involve proposals that an interlocutor is, implicitly or not, invited to answer. The

presence of a first part in an adjacency pair (Sacks et al. 1974) makes the second part

relevant. Lerner discusses this form of tacit address in certain action sequences: “Action

sequences can play an important part in the selection of a next speaker even when they

are not accompanied by an explicit form of addressing. The organization of actions - as

sequences of actions - shapes participation; each course of action shapes the

opportunities to participate within it. For example, asking a question ordinarily makes

an answer especially relevant.” (Lerner 2003: 190). Similarly, in the conversations

studied, questions and proposals always invite an answer or a confirmation. This can be

done through direct address, but is often achieved implicitly. Obviously, in the push-to-

talk conversations, the explicit marker is the pressing of the “Enter” key by the current

speaker, signalling the end of his or her turn.

Since strategies for turn-taking are different in each type of corpus, I have divided the

strategies in three types, basically the three types of actions that can happen in any

conversation. A speaker may (i) yield the turn; or (ii) hold the turn; and an interlocutor

may (iii) take the turn. I examine each in the following sections, and for each corpus

type.

4.1. Characteristics of turn yielding

Turn yielding is the most interesting aspect of the three, since in the push-to-talk

conversations there is no possibility of taking the turn (the interlocutor may not take the

turn until the current speaker has pressed “Enter”). Turn-yielding, therefore, allows a

full comparison across the two types of corpora.

Turn yielding in the push-to-talk conversations is achieved, obviously, by

mechanically passing the channel to the interlocutor. But here we are also interested in

what happens in the cross-talk conversations when the floor is passed by the current


speaker. This is achieved in a variety of ways: Pauses, address terms, questions, and tag

questions.

Pauses are frequently used as the most obvious signal that the current speaker desires

to yield the turn. In (2), speaker FANS proposes a meeting, which projects a second part

(an acceptance or a rejection), but makes the desire to yield the turn clearer with a pause.

(2) Cross-talk

fans_05: oye. <quiero> quiero /ah/ (0.8) tener una conferencia contigo.

(0.55)

flxb_06: sí. cómo no. cuándo?

‘Listen. I want I want uh to have a meeting with you.’

‘Yes. Of course. When?’

In push-to-talk conversations, pauses are also present. In Example (3), speaker

FSNM yields the turn (in the second turn in the example) by pushing a button, but a

short pause precedes that pushing. It is, of course, difficult to decide whether the pause

is part of the usual turn-taking mechanism, or whether the speaker is simply taking a

few seconds to press the turn-yielding key (e.g., trying to find it; or pausing so that the

end of her turn is not lost in the transition).

(3) Push-to-talk

ffcs_07: /mm/ el lunes, la verdad que solo tengo una hora en la mañana. pero el martes en

la tarde, qué tal?

fsnm_08: pues tengo casi toda la tarde libre a partir de cómo, las doce y media. (1.56)

ffcs_09: (1.13) perfecto. ...

‘Mm Monday, the truth is that I only have one hour in the morning. But Tuesday in the

afternoon, how’s that?’

‘I have almost all afternoon free after about twelve thirty.’

‘Perfect.’

Pauses were only transcribed if they were at least 0.2 seconds long. Ford and

Thompson (1996) report that a length of at least 0.3 seconds is relevant in turn-taking.

Therefore, pauses studied here are only those that could be involved in turn-taking.

Pause duration was measured using the Praat program, with an additional script that

342 Maite Taboada

marks pauses in speech5. In push-to-talk, pauses at the end of a turn indicate that there

was a silence before the speaker pressed the Enter key.

In Table 4, I summarize the presence of pauses in the conversations. We can see that

they are much more numerous in push-to-talk, especially given that there are fewer turns

in those conversations. There is almost one pause per turn, as compared to 27 pauses

overall in the 931 turns for the cross-talk data. Striking also is the presence of pauses at

turn boundaries. A large number happens at the end of turn, 31.45% of all the pauses

present in the push-to-talk data. Presumably, the speaker finishes his or her utterance,

and allows for the interlocutor to hear it fully (thus producing a pause) before turning

the floor over. However, in some cases, it seems that the interlocutor is not quite ready

to talk, as there are also pauses at the beginning of turns. On the other hand, pauses at

turn boundaries in cross-talk are minimal, and more frequent at the beginning of a turn,

where a speaker may take some time to answer a question that his or her interlocutor has

asked.

Push-to-talk Cross-talk

Total number of pauses 237 27

At the end of a turn 78 (31.45%) 1 (0.11%)

At the beginning of a turn 36 (14.52%) 6 (0.64%)

Total number of turns 248 931

Table 4. Pauses and turn-taking

Pauses in cross-talk happen most often turn-medially, when speakers provide a

chance for their interlocutor to take the floor, which is not taken up. Hopper (1992: 107)

describes turn-taking in telephone conversations - which are very similar to the cross-

talk conversations analyzed here - as a relay race. A pause is like a runner dropping the

baton. “If the runners drop the baton while it is being passed to the next runner, that next

runner should retrieve it. If the drop occurs away from such a transition place, the

current runner must retrieve it.” Such a situation is presented in (4). Speaker FLXB

replies to a suggestion of the 22nd with a ‘yes’ (after a pause), and a repetition of the

date. She then produces a pause, presumably to yield the floor. But the interlocutor does

not take up that opportunity, and FLXB repeats the date, making it more explicit that she

is free on that day. In the next turn, we understand why speaker FANS did not take the

turn at that pause: She wanted more specific information about when on that day FLXB

is available.

5 Praat is freely available: http://www.fon.hum.uva.nl/praat/. The script can also be downloaded

from the following web page: http://www.helsinki.fi/~lennes/praat-scripts/. Many thanks to Mietta Lennes

for the script.


(4) Cross-talk

fans_11: … qué te parece el veintidós?

flxb_12: (0.65) sí. el veintidós. (0.88) el veintidós está bien. estoy libre.

fans_13: todo el día?

‘…What do you think about the 22nd?’

‘Yes. The 22nd. The 22nd is good. I am free.’

‘All day?’

Address terms are used in very few cases. The speakers may, or may not, have known

each other before they came to record the conversations (they were recruited and

brought to a lab), but in all cases were introduced to each other, and thus know each

other’s names. In the following example, the speaker uses a combination of a direct

question and her interlocutor’s name to yield the turn. Address terms are not necessary,

since there is only one interlocutor, and that could explain their scarcity6.

(5) Push-to-talk

fmcs_07: … así que /mm/ qué te parece si ya quedamos en juntarnos a la una de la tarde

para almorzar, /eh/ no sé donde vos digas, y ya después /eh/ nos queda toda la tarde libre

para terminar el proyecto. te parece bien, Miriam?

‘… So mm what do you think if we arrange to meet then at 1 p.m. to have lunch, uh I don’t

know, wherever you want, and then we have the whole afternoon free to finish the project.

What do you think, Miriam?’

Direct questions are closely related to address terms. Given that the conversations

always involve two people, a direct question is addressed to the only other interlocutor.

Thus, even if a question contains no address term, the address term is implied. Example

(6) shows a direct address, with the syntactic structure of a question. In fact, speaker

MINM realizes early on that this is a question, and does not allow speaker MRRC to

finish his utterance, but overlaps as soon as the date (Friday the 23rd) has been

mentioned. Example (7) shows a push-to-talk instance of the same phenomenon, with a

question (which is actually not answered in the following turn).

(6) Cross-talk

mrrc_09: este, (0.5) qué te parecería, (0.33) <el> (0.54) el viernes veintitrés [ de julio. ]

6 Lerner (2003) also found that address terms are rare, even in multi-party conversation.

344 Maite Taboada

minm_10: [ viernes veintitrés, ] sería perfecto.

‘Uh, what about Friday <the> the 23rd [ of July. ]’

‘[ Friday the 23rd, ] would be perfect.’

(7) Push-to-talk

mjmg_07: ... el día treinta y uno, (0.95) tengo una clase de dos a cuatro. a ver cómo te va a

ti?

msnc_08: bueno. me podrías haber avisado antes. no? ...

‘… On the 31st, I have a class from two to four. Let’s see how’s that with you?’

‘Okay. You could have told me before. No?...’

Items projecting a second part in an adjacency pair are not always questions.

Sometimes they are suggestions that project an acceptance or a rejection. In (8), the first

speaker makes a suggestion to meet at the same place in future occasions. Syntactically,

the utterance is a statement, but it finishes with rising intonation. The other speaker

realizes the need for an answer, and produces one promptly, although he hesitates after

the initial ‘yes’, which leads to an instance of overlap.

(8) Cross-talk

fmgg_08: okay. perfecto. <perf> muy bien. <y siempre podríamos> /ah/ las próximas

reuniones siempre nos podríamos encontrar en el mismo lugar?

mmxb_09: sí. sí. <de> [ exactamente. ]

fmgg_10: [ /eh/ okay. ]

‘Okay. Perfect. Perf very good. And <we could always> uh the next meetings we could

always meet at the same place?’

‘Yes. Yes. Of exactly.

‘Uh okay.’

Questions may also take the form of tag questions, where confirmation, rather than

information, is requested. Sacks et al. (1974: 719) characterize tag questions as exit

devices for a turn, or post-completers. They indicate that the turn is complete, and the

interlocutor may take the turn. In Example (9), speaker MENM asks for confirmation of

the date proposed with a no?, a common tag question in Spanish.

(9) Cross-talk

menm_08: bueno. /eh/ mira. vamos <a> a tratar de encontrar aquí. qué te parece el día

nueve?


mgbm_09: el día nueve de febrero?

menm_10: sí. estamos a dos días de hoy, no?

mgbm_11: /uh_huh/. …

‘Okay. Uh look, let’s let’s try to meet here. How about the 9th?’

‘The 9th of February?’

‘Yes. That is two days from today, right?’

‘Uh-huh. …’

In general, first parts in adjacency pairs seem to be the most common turn-yielding

device. As mentioned above, they can be questions or statements asking about or

suggesting a meeting time, with a preferred second part being the acceptance of that

date, and a rejection as a dispreferred second part. Other pairs include greeting-greeting,

goodbye-goodbye, confirmation-confirmation. Table

5 displays the number of first parts in each type of conversation, broken down into

questions, tag questions and other. In push-to-talk, the majority of turns (81.85%)

involve the initiation of an adjacency pair. In cross-talk, the number is lower (58.97%),

but it still accounts for a majority of the turns. That is, speakers yield the turn because

they have initiated an adjacency pair that requires a second part, to be produced by the

addressee.


Questions 119 311

Tag questions 4 20

Other first parts 80 218

Total first parts 203 (81.85%) 549 (58.97%)

Total number of turns 248 931

Table 5. Adjacency pairs

Naturally, turns do not consist only of one part in an adjacency pair. Most often a

second part starts the turn and a first part finishes it. That is, the sequence is: Speaker

A’s proposal - Speaker B’s rejection + Speaker B’s new proposal. Elsewhere I have

extensively described the structure of the conversations in those terms (Taboada 2003,

2004). In (10) we can see a longer example of this succession of adjacency pairs.

Speaker FSNM produces a first part, a suggestion of a time to meet. FFCS produces a

dispreferred second part, rejecting that date (and giving a reason). She then continues

her turn with another first part, a new proposal to meet the following week. FSNM does

not directly give a second part to the new proposal, but indirectly rejects it by insisting

346 Maite Taboada

on meeting on the current week (a new first part). FSNM yields the turn (mechanically,

since this is a push-to-talk conversation), and FFCS answers the question with a

negative, but does not propose a new date, since she already had suggested moving the

date. In fact, turns 4 and 5 are a side sequence, and speaker FFCS’s proposal to meet the

following week receives a second pair in turn 6. It appears that speakers were keeping

track of the side sequence, since it is only turns 4 and 5 that contain a single part of the

pair.

(10) Push-to-talk

fsnm_02: bueno, el lunes para mí será bien como, a la hora de almorzar entre las once y

media, por allí, hasta la una y media podremos.

ffcs_03: no. el lunes no me queda bien. tengo un almuerzo de las doce a las dos. qué te

parece si <lo hac> lo dejamos para el próximo lunes.

fsnm_04: no tienes ninguna hora entre martes y viernes de esa semana?

ffcs_05: no. no tengo. este <tengo> me voy de viaje por tres días y el viernes tengo clase y

dos reuniones.

fsnm_06: bueno, el lunes, por la mañana será muy bueno para mí pero por la tarde tengo

una reunión entre las dos y las cuatro.

‘Okay, Monday would be good for me, around lunchtime between 11:30, or so, until 1:30

we could.’

‘No. Monday is not good for me. I have a lunch between 12 and 2. What do you think if we

<do it> leave it until next Monday.’

‘Don’t you have any time between Tuesday and Friday this week?.’

‘No. I don’t. Uh <I have> I’m going away for three days and Friday I have a class and two

meetings.’

‘Okay, on Monday, in the morning would be very good for me, but in the afternoon I have a

meeting between 2 and 4.’

The comparison between the two types of conversations shows that silences are

proportionately more common in push-to-talk conversations, and that in both types turn-

taking seems to follow the structure of adjacency pairs. It is interesting that in the push-

to-talk data, a large number of the turns (81.85%) are first parts in an adjacency pair,

whereas in cross-talk, although still the majority, first parts are present in 58.97% of the

turns. It seems that push-to-talk conversations are more focused on the task, and more

compact: Most turns contain a task-related element (a first part).


4.2. Characteristics of turn holding

In conversation in general, and in task-oriented conversation in particular, speakers may

pass the turn to an interlocutor when that interlocutor is not yet ready to take it. As we

saw in the previous section, a speaker may utter a first part in an adjacency pair, through

a question, a command, or a mention of time availability. Then it is clear that the

interlocutor has to answer or otherwise respond to the offer. This happens in both types

of conversations. In the push-to-talk data, the passing of the turn is very clear, because

the speaker pushes a button that opens the channel to the interlocutor. The other speaker

may wish to accept that passing of the turn, but may not be ready to provide a full

answer yet. Then he or she holds the floor through a number of devices: Silent pauses,

filled pauses, and discourse markers.

Silent pauses, that is, periods of time when nobody talks, are the least effective

method of holding the floor. A pause may indicate a number of things, among them that

the communication has broken down and needs to be repaired. When the turn is passed

by a speaker, and the interlocutor produces a pause, the speaker that passed the turn may

feel that something went wrong, and reclaim the turn, maybe providing clarification. Or

he or she may not say anything, waiting for the interlocutor to speak. We saw, in Table

4, that pauses at the beginning of a turn are rare: 14.52% of all the pauses in push-to-talk

data happened at the beginning of a turn. Recall that in those conversations, the current

speaker may hold the floor indefinitely. However, they did not pause very often. One

example is given in (11), where speaker MJMG proposes a date, and passes the turn.

Speaker MSNC pauses for 1.42 seconds before he narrows down his availability for that

date.

(11) Push-to-talk

mjmg_01: (0.96) bueno. vamos a ver. (0.46) /eh/ qué tal el veinticuatro? (0.92) yo puedo

cualquier momento, menos de una, a cuatro. (0.67)

msnc_02: (1.42) el veinticuatro yo podría, pero tengo una reunión, desde las diez hasta las

doce. tú podrías de diez a doce? (1.12)

‘Okay. Let’s see. How about the 24th? I can any time, except for from 1, to 4.’

‘On the 24th I could, but I have a meeting, from 10 until 12. Could you from 10 to 12?’

Pauses at the beginning of the turn are much less frequent in cross-talk conversation:

6 out of a total of 27 pauses (0.64%) happened at the beginning of a turn, as we saw in

the previous section. For example, in (12), speaker FANS pauses at the beginning of her

turn (turn number 15), presumably because she is checking her schedule for the day

proposed.

(12) Cross-talk:

348 Maite Taboada

flxb_12: (0.65) sí. el veintidós. (0.88) el veintidós está bien. estoy libre.

fans_13: todo el día?

flxb_14: todo el día.

fans_15: (0.54) oh yo también. qué bien.

‘Yes. The 22nd. The 22nd is good. I’m free.’

‘All day?’

‘All day.’

‘Oh, me too. That’s great.’

A filled pause indicates more clearly that the interlocutor, to whom the floor has just

been passed, wishes to talk, but is not quite ready to do so. Filled pauses take a number

of forms: eh, ah, mm, uh. They rarely appear alone, rather being accompanied by a

pause, a discourse marker, or both. Example (13) shows a filled pause on its own, eh,

not only at the beginning of his turn, but also throughout the turn.

(13) Push-to-talk

mrnn_05: /eh/ yo paso por tu oficina /eh/ a las dos de la tarde, <b> porque tengo que /eh/

volver <a c> a casa después de la clase que tengo <en> en la mañana, …

‘Uh I’ll come by your office uh at two p.m., <b> because I have to uh come back <h> home

after the class that I have <in> in the morning, …’

Discourse markers are the most frequent turn-holding device. The definition of

discourse marker that I have taken is quite broad. It includes the following: A ver,

vamos a ver (‘let’s see’), oye, oyes (‘listen’), mira, mirá (‘look’), ya (‘ok’), déjame ver

(‘let me see’), pues (‘well/then’), ay (‘uh oh’), este (literally, ‘this’), and the English ok.

There are also combinations of them, such as mira oye, a ver pues, or ya ok. The

distinction between a discourse marker and a filled pause is not always clear. The

method that I followed was to classify non-words as filled pauses, and any other

adverbs, verbs and conjunctions, as discourse markers.

Byron and Heeman (1998) suggest that discourse markers are more prevalent in task-

oriented spoken dialogue: in the TRAINS corpus of task-oriented dialogues that they

analyzed, 44.1% of the turns were introduced with a discourse marker (that figure

excludes acknowledgements and filled pauses). They found a number of functions were

realized by discourse markers, among them acknowledgements, repairs, and signals

about the type of conversational move about to be produced. For example, utterances

that summarize previous contributions start with so, and utterances that express dissent

(a dispreferred second pair) start with well. They also found that (preferred) second parts

of adjacency pairs did not usually start with a discourse marker. They conclude that

discourse markers are used most frequently when there are no strong expectations about


the utterance that the speaker is about to make. This seems to be the case in the corpora

that I am describing here: Discourse markers are used, sometimes in combination with

filled pauses, when the speaker him or herself is not sure about what move they need to

make next: They need to consider whether they will accept or reject a proposal; or

whether they can make a new proposal for a meeting date. An example is presented in

(14), where the speaker uses a discourse marker (la verdad, ‘the truth’), plus a filled

pause (eh) to reject the date just proposed by the other speaker.

(14) Push-to-talk

fmbo_01: … yo estoy un poco ocupada esta semana porque me voy de viaje. pero (0.43)

qué le parece la semana que viene /eh/, <el trece?> el martes trece, a la mañana .

menc_02: la verdad la verdad, /eh/ el trece no es el mejor día para reunirnos para mí.

‘… I’m a bit busy this week because I’m going away. But what do you think about next

week uh, the 13th? Tuesday the 13th, in the morning.’

‘The truth the truth, uh the 13th is not the best day to meet for me.’

In other cases, the discourse marker introduces a request for a clarification, in itself

also a dispreferred second part, as in Example (15), from the cross-talk corpus, where

both oh and a ver (‘let’s see’) give the speaker some time to think about the next move:

FVNM is not ready to say “yes” to going to a movie until she knows what day is

proposed, and as a consequence she requests further information.

(15) Cross-talk

fknh_03: te hablo para decirte de una cita. si quieres ir conmigo al cine.

fvnm_04: oh, a ver. qué día es?

‘I’m talking to you to tell you about a meeting. If you want to go with me to the movies.’

‘Oh, let’s see. What day?’

The discourse marker bueno deserves special attention. It is the most common

marker, by far, in both types of conversation. Out of the 80 discourse markers present at

the beginning of the turn in push-to-talk, 48 are instances of bueno (60%). In cross-talk,

bueno amounts to 53 instances out of a total of 166 discourse markers (32%). Bueno is

used in isolation, or together with another marker, with a filled pause, or with a pause

(ah bueno, oye bueno). It is used to signal a dispreferred second part in an adjacency

pair (Cortés Rodríguez 1998), a use similar to that of well in English (Schiffrin 1987).

Example (16) illustrates such use of bueno: Speaker FLNW issues an implicit blanket

rejection of all the dates previously proposed by speaker MPNE, and suggests to look at

a different week. The rejection is implicit in the bueno, and the new proposal seems to

be foreshadowed by mira (‘look’).

350 Maite Taboada

(16) Push-to-talk

mpne_01: /ehm/ <la> la idea es que como te digo el lunes, en la mañana, martes /eh/ /um/

<de> en la tarde, el miércoles a cualquier hora, y el jueves en la mañana.

flnw_02: bueno. mira. yo creo que vamos a tener que revisar para la semana siguiente

porque por lo que conversamos, nuestras agendas están totalmente encontradas. …

‘Um <the> the idea is that, as I said, Monday, in the morning, Tuesday uh um <of> in the

afternoon, Wednesday anytime, and Thursday in the morning.’

‘Well. Look. I think we are going to have to revise for next week because from what we’ve

talked about, our schedules are completely in conflict….’

Example (17) is another instance of bueno as a turn-holder. Speaker FMSG has

rejected a number of proposals, including the most recent (Friday), and passes the turn.

Speaker FDXH does not seem ready to put forth a new proposal, and hesitates with a

number of filled pauses, including no, something that sounds close to the French oh là

là, and then a bueno, followed by another filled pause (ah), and two new markers

combined (pues mira), until she finally finds her next available slot (the Wednesday

after).

(17) Cross-talk

fmsg_07: …no hombre. se va a poner imposible. el viernes /begin_laugh/ no voy a

/end_laugh/ poder. no tengo más que cuatro a cinco. no nos alcanza el tiempo.

fdxh_08: no. <n> ah la lá bueno. ah pues mira. yo no tengo hasta el otro miércoles…

‘… No man. It’s looking impossible. On Friday I won’t be able to. I only have from 4 to 5.

That’s not enough time.’

‘No <n> ah la la well. Ah, well look. I won’t have (anything) until the Wednesday after…’

The marker bueno has a few other uses not directly related to turn-taking. It can

indicate the acceptance of a proposal, and therefore the closing of the conversation, as

shown in (18), where it is more appropriately translated as ‘okay’, instead of ‘well’, as

we have seen in the previous examples.

(18) Push-to-talk

ffcs_07: /mm/ el lunes, la verdad que solo tengo una hora en la mañana. pero el martes en

la tarde, qué tal?

fsnm_08: pues tengo casi toda la tarde libre a partir de como, las doce y media. (1.56)

ffcs_09: (1.13) perfecto. porque no nos reunimos <a la una> /ah/ de una a las tres. qué tal?

(1.7)


fsnm_10: (0.92) será muy bueno esa hora, entonces te veré allí.

ffcs_11: bueno. está bien. te veo el martes a la una. hasta luego.

‘Mm, on Monday the fact is that I only have one hour in the morning. But Tuesday in the

afternoon, how’s that?’

‘Well, I have almost all afternoon free, from about 12:30 on.’

‘Perfect. Why don’t we meet <at one> uh from one to three. How’s that?’

‘It’ll be very good at that time, then I’ll see you there.’

‘Okay. That’s good. I’ll see you on Tuesday at one. See you later.’

Repetitions also serve as turn-holders. In (19), speaker FANS repeats, in the last turn

of the example, the date just proposed (the 17th), with rising intonation. This could be

both a request for a confirmation and a turn holder. The request for confirmation

hypothesis is supported by the fact that there is a pause after el diecisiete?, but the turn

holder hypothesis is possible, since FANS continues to talk without having received

confirmation. She may just have said something to hold the turn, and to indicate that she

is considering that date.

(19) Cross-talk:

fans_09: y a ver. qué otro día yo puedo.

flxb_10: (0.64) el diecisiete puedes?

fans_11: el diecisiete? (1.59) no puedo el diecisiete. (1.8) <qu> qué te parece el veintidós?

‘And let’s see what other day I can.’

‘Can you on the 17th?’

‘The 17th? I can’t on the 17th. <Wh> What do you think of the 22nd?’

Pauses, filled pauses, and discourse markers are also used in combination. In

Example (20), speaker MFMM starts with a marker (a ver), and then continues with a

filled pause and an address term, before pointing out a problem with the time just

proposed by his interlocutor.

(20) Push-to-talk

mfmm_08: a ver, /eh/ Octavio. creo que las dos horas tienen que ser seguidas. va a ser una

reunión bastante larga, tiene que ser de dos horas y las dos horas seguidas. …

‘Let’s see, uh Octavio. I think that the two hours have to be contiguous. It’s going to be a

pretty long meeting, it has to be two hours and the two hours contiguous. …’

352 Maite Taboada

Table 6 summarizes the numbers of filled pauses and discourse markers in each of

the corpora, and their averages per conversation. This represents only instances that

occur at the beginning of the turn.


Filled pauses 24 108

Discourse markers 80 166

Filled pauses per conversation 0.8 3.6

Discourse markers per conversation 2.6 5.5

Table 6. Filled pauses and discourse markers

Although it is difficult to compare the numbers (given that cross-talk conversations

contain more turns and more words), we can conclude that, on average, each push-to-

talk conversation contained fewer filled pauses and fewer discourse markers. In other

words, the speakers completed the same task using fewer filled pauses and discourse

markers in the push-to-talk setting. One conclusion of this data is that those features are

necessary when managing spontaneous conversation, but they drop in frequency once

the pressure of holding the turn disappears.

4.3. Characteristics of turn taking

Turn-taking in the push-to-talk conversations does not happen freely, but is controlled

by the interlocutor: A speaker can take the turn only if the interlocutor yields it. Once

the turn has been passed, we are in a turn-holding situation. It is only in the cross-talk

conversations that we can observe actual instances of turn taking, that is, a speaker

taking the floor from the current speaker. In Sacks et al.’s (1974) terms, the speaker is

self-selecting at the transition-relevance place. Since there is no possible comparison in

this respect between the two corpora, I will not have much to say about how turn-taking

happens, apart from some observations about overlapping.

Overlapping is the most obvious instance of self-selection by an interlocutor, since it

indicates that the current speaker is not ready to yield the turn. In the cross-talk data,

22% of the turns contain some overlapping or simultaneous talk (204 out of 931 turns).

In most cases, the simultaneous talk is clearly not an instance of trying to take the floor,

but merely a backchannel, as in (21), where speaker MLGZ produces a backchannel that

signals understanding that Monday is not good (ah), but then lets speaker MLPM finish

his turn before he asks the following question (‘and when are you back’).


(21) Cross-talk

mlpm_03: este el lunes estoy de viaje. [ el lunes, ] <me> muy complicado para mí.

mlgz_04: [ ah. ] y cuándo volvés?

‘Eh Monday I’m away. [ Monday, ] <me> very busy for me.’

‘[ ah ] and when are you back?’

In other cases, it is more plausible to think that the interlocutor was trying to take the

floor, sometimes to preclude discussion of a date, or to make the interlocutor stop when

a date has been proposed, so that the self-selecting speaker gets the chance to consult his

or her calendar. In (22), speaker FDXH suggests Friday, and ends her question with

rising intonation. At that point, speaker FMSG repeats ‘Friday’, also with rising

intonation, ready to take the floor and check her schedule. But speaker FDXH has not

finished her turn, and repeats the Friday proposal, specifying that she is free then. The

overlapping talk happens at the point where FDXH repeats el viernes, because FMSG

was prompt to take the floor at that point. In fact, FMSG could naturally have expected

her interlocutor’s turn to end at the point where FDXH produced a question. According

to Schegloff (1988: 141), “if a turn has several components (that is, turn-constructional

units) in it, one of which is a question, the question is almost always the last of them, for

on its completion, the question will ordinarily have made it someone else’s turn to

talk.”.

(22) Cross-talk

fdxh_04: bueno. qué te parece /eh/ <ma> el viernes? [ el viernes ] tengo libre.

fmsg_05: [ el viernes? ] el viernes. ah caray. déjame ver. este, tengo una reunión. con este

David ay ya no me acuerdo cómo se apellida. /begin_laugh/ <d> de /end_laugh/ diez a

doce.

‘Okay. What do you think uh <ma> on Friday? [ Friday ] I’m free.’

‘[ Friday? ] Friday. Oh jeez. Let me see. Um, I have a meeting. With this David, uh I don’t

remember his last name. <f> from ten to twelve.’

5. Discussion

I have presented an analysis of turn-taking strategies in two different sets of task-

oriented conversations. The comparison between the two is informative because they

were recorded in exactly the same circumstances, with the exception of the turn-taking

354 Maite Taboada

mechanisms: Speakers in one group of conversations controlled the turn mechanically,

resulting in one-way communication.

The conversations have been examined in terms of three characteristics: Turn

yielding, turn holding and turn taking. The two types are different in terms of turn

yielding and turn holding. In both cases, pauses play a role. Pauses are more frequent at

the ends and beginnings of turns in the push-to-talk data, maybe a reflection of the

mechanical intervention. On the other hand, cross-talk, that is, spontaneous

conversations, have a higher number of filled pauses, probably as a result of the need to

show that the turn is not being yielded.

The most significant differences overall are, in push-to-talk, the lower numbers of

pauses, filled pauses and discourse markers; in cross-talk, the presence of overlap. As it

has been pointed out before, those are common characteristics of spontaneous

conversation. In general, pauses, if present, are very brief, and overlap exists, but it is

also brief. Filled pauses and discourse markers are used to manage the interaction, and

to hold the turn. When we control the turn-taking through mechanical means, the typical

features of turn-taking decrease in frequency.

The conclusions that we can draw from this data are potentially limited to the data

itself. Neither type of conversation was completely spontaneous. Even the cross-talk

data was recorded in a lab, in somewhat artificial conditions. Most conversation analysis

practitioners would frown upon such data. I feel that we cannot discard data because it

does not conform to some ideal of what “spontaneous” means. At the very least, we can

conclude that these are perfect instances of conversations produced in a particular

setting: The recording laboratory. Since we have a large number of conversations, and

since they were all recorded under similar circumstances, I believe that generalizations

about their characteristics are possible.

The research into turn-taking signals is useful in itself: We want to understand how

conversational participants signal to each other that they desire to yield or take the turn.

But it has many other applications, one of them the design of intelligent conversational

agents. Researchers in Artificial Intelligence have been designing systems that can

interact with a user, provide directions or information. One example is a conversational

agent developed at MIT that provides a virtual tour of real estate for sale. Bickmore and

Cassell (2005) found that the virtual real estate agent needs to know when and how to

yield and take the turn, using not only verbal cues, but also gesture and posture. The

nonverbal cues were also important in building trust with the virtual agent.


References

Auer, Peter (1996) On the prosody and syntax of turn-continuations. In E. Couper-Kuhlen and M. Selting

(eds.), Prosody in Conversation: Interactional Studies. Cambridge: Cambridge University Press, pp. 57-

100.

Bangerter, Adrian, Herbert H. Clark, and Anna R. Katz (2004) Navigating joint projects in telephone

conversations. Discourse Processes 37.1: 1-23.

Beattie, Geoffrey (1977) The dynamics of interruption and the filled pause. The British Journal of Social

and Clinical Psychology 16.3: 283-284.

Beattie, Geoffrey (1979) Planning units in spontaneous speech: Some evidence from hesitation in speech

and speaker gaze direction in conversation. Linguistics 17: 61-78.

Beattie, Geoffrey (1981) The regulation of speaker turns in face-to-face conversation: Some implications

for conversation in sound-only communication channels. Semiotica 34.1-2: 55-70.

Beattie, Geoffrey, Anne Cutler, and Mark Pearson (1982) Why is Mrs. Thatcher interrupted so often?

Nature 300.23: 744-747.

Bickmore, Timothy W., and Justine Cassell (2005) Social dialogue with embodied conversational agents.

In J. van Kuppevelt, L. Dybkjaer and N. Bernsen (eds.), Advances in Natural Multimodal Dialogue

Systems. Berlin: Springer.

Briz, Antonio (1993) Los conectores pragmáticos en español coloquial (I): Su papel argumentativo.

Contextos 11.21-22: 145-188.

Brown, Penelope, and Stephen Levinson (1978) Universals in language use: Politeness phenomena. In E.

N. Goody (ed.), Questions and Politeness: Strategies in Social Interaction. Cambridge: Cambridge

University Press, pp. 56-289.

Byron, Donna K., and Peter A. Heeman (1998) Identifying discourse markers in spoken dialog, AAAI

Spring Symposium on Applying Machine Learning and Discourse Processing. Stanford, CA.

Chafe, Wallace (1994) Discourse, Consciousness and Time: The Flow and Displacement of Conscious

Experience in Speaking and Writing. Chicago: University of Chicago Press.

Condon, Sherri L. (1986) The discourse functions of OK. Semiotica 60: 73-101.

356 Maite Taboada

Condon, Sherri L. (2001) Discourse ok revisited: Default organization in verbal interaction. Journal of

Pragmatics 33: 491-513.

Cortés Rodríguez, Luis (1998) Marcadores del discurso y análisis cuantitativo. In M.A. Martín Zorraquino

and E. Montolío Durán (eds.), Los marcadores del discurso: Teoría y análisis. Madrid: Arco, pp. 143-

160.

Cutler, Anne, and Mark Pearson (1986) On the analysis of prosodic turn-taking cues. In C. Johns-Lewis

(ed.), Intonation in Discourse. San Diego, CA: College Hill, pp. 139-156.

Duncan, Starkey (1972) Some signals and rules for taking speaking turns in conversations. Journal of

Personality and Social Psychology 23: 283-292.

Duncan, Starkey (1973) Toward a grammar for dyadic conversation. Semiotica 9.1: 29-46.

Duncan, Starkey, and Donald W. Fiske (1977) Face-to-Face Interaction: Research, Methods, and

Theory. Hillsdale, NJ: Lawrence Erlbaum.

Duncan, Starkey, and Donald W. Fiske (1985) Interaction Structure and Strategy. Cambridge: Cambridge

University Press.

Edelsky, Carole (1981) Who's got the floor? Language in Society 10: 383-421.

Ferrara, Kathleen (1997) Form and function of the discourse marker 'anyway': Implications for discourse

analysis. Linguistics 35: 343-378.

Ford, Cecilia, Barbara A. Fox, and Sandra A. Thompson (1996) Practices in the construction of turns: The

'TCU' revisited. Pragmatics 6.3: 427-454.

Ford, Cecilia, and Sandra A. Thompson (1996) Interactional units in conversation: Syntactic, intonational,

and pragmatic resources for the management of turns. In E. Ochs, E. Schegloff and S.A. Thompson (eds.),

Interaction and Grammar. Cambridge: Cambridge University Press, pp. 134-184.

Fox Tree, Jean E., and Josef C. Schrock (1999) Discourse markers in spontaneous speech: Oh what a

difference an oh makes. Journal of Memory and Language 40: 280-295.

Fox Tree, Jean E., and Josef C. Schrock (2002) Basic meanings of you know and I mean. Journal of

Pragmatics 34: 727-747.

Fraser, Bruce (1999) What are discourse markers? Journal of Pragmatics 31: 931-952.


Furo, Hiroko (2001) Turn-Taking in English and Japanese: Projectability in Grammar, Intonation, and

Semantics. New York: Routledge.

Goffman, Erving (1981) Forms of Talk. Pennsylvania: University of Pennsylvania Press.

Goodwin, Charles (1981) Conversational Organization: Interaction between Speakers and Hearers. New

York: Academic Press.

Grosz, Barbara J., and Candace L. Sidner (1986) Attention, intentions, and the structure of discourse.

Computational Linguistics 12.3: 175-204.

Hayashi, R. (1991) Floor structure of English and Japanese conversation. Journal of Pragmatics 16: 1-30.

Heritage, John (1984) A change-of-state token and aspects of its sequential placement. In J.M. Atkinson

and J. Heritage (eds.), Structures of Social Action: Studies in Conversation Analysis. Cambridge:

Cambridge University Press, pp. 299-345.

Hidalgo, Antonio (1998) Alternancia de turnos y conversación. Sobre el papel regulador de los

suprasegmentos en el habla simultánea. Lingüística Española Actual 20.2: 217-238.

Hopper, Robert (1992) Telephone Conversation. Bloomington, IN: Indiana University Press.

Houtkoop, Hanneke, and Harrie Mazeland (1985) Turns and discourse units in everyday conversation.

Journal of Pragmatics 9: 595-619.

Jaffe, Joseph, and Stanley Feldstein (1970) Rhythms of Dialogue. New York: Academic Press.

Jefferson, Gail (1983) On a failed hypothesis: 'Conjunctionals' as overlap-vulnerable. Tilburg Papers in

Language and Literature 28: 29-33.

Jefferson, Gail (1984) Notes on a systematic deployment of the acknowledgment tokens "yeah" and "mm

hm". Papers in Linguistics 17: 197-216.

Jones, Rod, and Joanna Thornborrow (2004) Floors, talk and the organization of classroom activities.

Language in Society 33: 399-423.

Kendon, Adam (1994) Do gestures communicate?: A review. Research on Language and Social

Interaction 27.3: 175-200.

358 Maite Taboada

Kendon, Adam (1995) Gestures as illocutionary and discourse structure markers in Southern Italian

conversation. Journal of Pragmatics 23.3: 247-279.

Kendon, Adam (2002) Some uses of the head shake. Gesture 2.2: 147-182.

Knott, Alistair, and Ted Sanders (1998) The classification of coherence relations and their linguistic

markers: An exploration of two languages. Journal of Pragmatics 30: 135-175.

Lerner, Gene H. (2003) Selecting next speaker: The context-sensitive operation of a context-free

organization. Language in Society 32.2: 177-201.

Local, John (1996) Conversational phonetics: Some aspects of news receipts in everyday talk. In E.

Couper-Kuhlen and M. Selting (eds.), Prosody in conversation. Cambridge: Cambridge University Press,

pp. 177-230.

Local, John, and John Kelly (1986) Projection and "silences": Notes on phonetic and conversational

structure. Human Studies 9: 185-204.

Maclay, Howard, and Charles E. Oswood (1959) Hesitation phenomena in spontaneous English speech.

Word 15: 19-44.

Martin, James R. (1984) Language, register and genre. In F. Christie (ed.), Children Writing: Reader.

Geelong, Victoria: Deakin University Press, pp. 21-30.

Pierrehumbert, Janet (1980) The Phonology and Phonetics of English intonation. Ph.D. dissertation, MIT,

Cambridge, Mass.

Placencia, María Elena (1997) Opening up closings---the Ecuadorian way. Text 17.1: 53-81.

Pomerantz, Anita (1984) Agreeing and disagreeing with assessments: Some features of

preferred/dispreferred turn shapes. In J.M. Atkinson and J. Heritage (eds.), Structures of Social

Interaction: Studies in Conversation Analysis. Cambridge: Cambridge University Press, pp. 57-101.

Raymond, Geoffrey (2004) Prompting action: The stand-alone "so" in ordinary conversation. Research on

Language and Social Interaction 37.2: 185-218.

Redeker, Gisela (1990) Ideational and pragmatic markers of discourse structure. Journal of Pragmatics

14: 367-381.

Redeker, Gisela (1991) Review article: Linguistic markers of linguistic structure. Linguistics 29.6: 1139-

1172.


Sacks, Harvey (1992) Lectures on Conversation (Gail Jefferson ed. Vol. II). London: Sage.

Sacks, Harvey, Emmanuel Schegloff, and Gail Jefferson (1974) A simplest systematics for the

organization of turn-taking in conversation. Language 50: 696-735.

Schegloff, Emmanuel (1968) Sequencing in conversational openings. American Anthropologist 70: 1075-

1095.

Schegloff, Emmanuel (1982) Discourse as an interactional achievement: Some uses of uh huh and other

things that come between sentences. In D. Tannen (ed.), Analyzing Discourse: Text and Talk. Georgetown

University Roundtable on Languages and Linguistics. Washington, DC: Georgetown University Press, pp.

71-93.

Schegloff, Emmanuel (1988) Discourse as an interactional achievement II: An exercise in Conversation

Analysis. In D. Tannen (ed.), Linguistics in Context: Connecting Observations and Understanding.

Norwood, NJ: Ablex, pp. 135-158.

Schegloff, Emmanuel, and Harvey Sacks (1973) Opening up closings. Semiotica 8: 289-327.

Schiffrin, Deborah (1987) Discourse Markers. Cambridge: Cambridge University Press.

Selting, Margret (1998) TCUs and TRPs: The construction of units in conversational talk. InLiSt -

Interaction and Linguistic Structures 4: http://inlist.uni-konstanz.de/issues/4/index.htm.

Selting, Margret (2000) The construction of units in conversational talk. Language in Society 29: 477-

517.

Serrano, María José (1995) El uso de la verdad y pues como marcadores discursivos de respuesta.

Español Actual 64: 5-16.

Stephens, Jane, and Geoffrey Beattie (1986) On judging the ends of speaker turns in conversation.

Journal of Language and Social Psychology 5.2: 119-134.

Taboada, Maite (2003) Modeling task-oriented dialogue. Computers and the Humanities 37.4: 431-454.

Taboada, Maite (2004) Building Coherence and Cohesion: Task-Oriented Dialogue in English and

Spanish. Amsterdam and Philadelphia: John Benjamins Publishing Company.

360 Maite Taboada

Taboada, Maite (in press) Discourse markers as signals (or not) of rhetorical relations. Journal of

Pragmatics.

Tanaka, Hiroko (2001) Adverbials for turn projection in Japanese: Toward a demystification of the

"telepathic" mode of communication. Language in Society 30.4: 559-587.

Wennerstrom, Ann, and Andrew F. Siegel (2003) Keeping the floor in multiparty conversations:

Intonation, syntax, and pause. Discourse Processes 36.2: 77-107.

Yngve, Victor H. (1970) On getting a word in edgewise. In Papers from the Sixth Regional Meeting of the

Chicago Linguistics Society. Chicago: University of Chicago, pp. 567-577.

Documents

Turn Taking Phrases